Data Anomaly Detection and Interpolation for Power Transmission Line Icing Monitoring based on Multi-modal Attention GAN

An increasing number of transmission lines now pass through remote forested areas, which are highly vulnerable to extreme weather and prone to ice accumulation. Failure to detect such ice accumulation in a timely manner can lead to severe consequences. Existing icing monitoring methods often lack consideration of the correlations among different modal data and are inefficient in predicting missing data and detecting anomalies, thereby reducing monitoring precision. To address these issues, we propose a data anomaly detection and interpolation method for power transmission line icing monitoring based on multi-modal attention GAN. Firstly, a multi-modal attention mechanism is constructed to assign weights to the input data. Secondly, the multi-modal attention weights and long short-term memory (LSTM) networks are employed to train the generator and discriminator of a generative adversarial network (GAN). The trained model is then utilized to detect anomalies in icing monitoring data, and segments with high anomaly scores are interpolated and modeled. Finally, simulations are conducted to validate the superior performance of the proposed algorithm in terms of precision, convergence, and robustness.